Process mining aims to analyze the business processes of organizations based on event logs. In the supply chain processes where multiple organizations are involved, it poses crucial challenges due to the information silo, privacy and interoperability concerns. This paper introduces a unified framework for vertical federated process mining, which addresses the challenges of discovering process models across multiple organizations without sharing raw sensitive data.
This study proposes a method for anonymizing local event logs and merging them into a federated event log, using time minimization strategies to correlate events and generate global case identifiers. The framework uses public datasets to compare the accuracy of the discovered process models against centralized and other approaches.
The results demonstrate that the proposed framework can reconstruct event logs and discover process models despite the existence of data format disparities and complex inter-organizational dependencies while respecting the privacy of the collaborating organizations. The proposed method achieved comparable performance to the centralized approach and outperformed the naïve approach.
It has a limitation on discovering vertical processes that have a lot of looping (cyclic) processes or cases that need to be solved outside of established procedures.
It aims to discover and analyze inter-organizational processes with privacy-sensitive information protection while integrating event data from multiple organizations' local event logs into a unified federated event log for a vertical federated setting.
This study addresses the need for a new approach in federated process mining due to rising privacy regulations. It provides a novel framework for vertical federated process mining, expanding the application of process mining techniques to inter-organizational contexts and offering a solution to maintain privacy while enabling process discovery.
